“Gesar”Thangka is an important type of Tibetan Tangka.With the generation and spread of Gesar epic,“Gesar”Thangka is also undergoing complex historical changes in the social and historical process of Tibetan areas...“Gesar”Thangka is an important type of Tibetan Tangka.With the generation and spread of Gesar epic,“Gesar”Thangka is also undergoing complex historical changes in the social and historical process of Tibetan areas.Since the 1980s,in the context of the rescue of Gesar epic,the rise of market economy,and the protection of intangible cultural heritage,“Gesar”Thangkar has gained creative vitality and experienced modern changes,presenting the aesthetic characteristics of secularization,sanctification,and art.展开更多
In October of 2006,"Treasures of the Chinese Nation,an Exhibition of Tibetan Culture"was held exclusively at the Capital Museum in Beijing.Over 200 exquisite Tibetan cultural relics and 100-odd Tibetan folkl...In October of 2006,"Treasures of the Chinese Nation,an Exhibition of Tibetan Culture"was held exclusively at the Capital Museum in Beijing.Over 200 exquisite Tibetan cultural relics and 100-odd Tibetan folklore images unveiled in the exhibition hall attracted endless lines of visitors.While monks were busy making Mandala,a painter was sketching Thangka in the hall.In conjunction with the traditional art show in the exhibition (?) the visitors were profoundly inspired by such unique demonstrations of Tibe...展开更多
当前大语言模型的兴起为自然语言处理、搜索引擎、生命科学研究等领域的研究者提供了新思路,但大语言模型存在资源消耗高、推理速度慢,难以在工业场景尤其是垂直领域应用等方面的缺点。针对这一问题,提出了一种多尺度卷积神经网络(convo...当前大语言模型的兴起为自然语言处理、搜索引擎、生命科学研究等领域的研究者提供了新思路,但大语言模型存在资源消耗高、推理速度慢,难以在工业场景尤其是垂直领域应用等方面的缺点。针对这一问题,提出了一种多尺度卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,LSTM)融合的唐卡问句分类模型,本文模型将数据的全局特征与局部特征进行融合实现唐卡问句分类任务,全局特征反映数据的本质特点,局部特征关注数据中易被忽视的部分,将二者以拼接的方式融合以丰富句子的特征表示。通过在Thangka数据集与THUCNews数据集上进行实验,结果表明,本文模型相较于Bert模型在精确度上略优,在训练时间上缩短了1/20,运算推理时间缩短了1/3。在公开数据集上的实验表明,本文模型在文本分类任务上也表现出了较好的适用性和有效性。展开更多
基金the National Social Science Foundation in 2017:stage achievement of"The Study of'Gesar'Thangka in the Inheritance of Tibetan Epic Culture"(Project Number 17BH153)the stage achievements of 2017 National Social Science Foundation major project"Investigation and Research on Image Culture of Hero Epic'Gesar'and Database Construction"(Project Number 17ZDA160).
文摘“Gesar”Thangka is an important type of Tibetan Tangka.With the generation and spread of Gesar epic,“Gesar”Thangka is also undergoing complex historical changes in the social and historical process of Tibetan areas.Since the 1980s,in the context of the rescue of Gesar epic,the rise of market economy,and the protection of intangible cultural heritage,“Gesar”Thangkar has gained creative vitality and experienced modern changes,presenting the aesthetic characteristics of secularization,sanctification,and art.
文摘In October of 2006,"Treasures of the Chinese Nation,an Exhibition of Tibetan Culture"was held exclusively at the Capital Museum in Beijing.Over 200 exquisite Tibetan cultural relics and 100-odd Tibetan folklore images unveiled in the exhibition hall attracted endless lines of visitors.While monks were busy making Mandala,a painter was sketching Thangka in the hall.In conjunction with the traditional art show in the exhibition (?) the visitors were profoundly inspired by such unique demonstrations of Tibe...
文摘当前大语言模型的兴起为自然语言处理、搜索引擎、生命科学研究等领域的研究者提供了新思路,但大语言模型存在资源消耗高、推理速度慢,难以在工业场景尤其是垂直领域应用等方面的缺点。针对这一问题,提出了一种多尺度卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,LSTM)融合的唐卡问句分类模型,本文模型将数据的全局特征与局部特征进行融合实现唐卡问句分类任务,全局特征反映数据的本质特点,局部特征关注数据中易被忽视的部分,将二者以拼接的方式融合以丰富句子的特征表示。通过在Thangka数据集与THUCNews数据集上进行实验,结果表明,本文模型相较于Bert模型在精确度上略优,在训练时间上缩短了1/20,运算推理时间缩短了1/3。在公开数据集上的实验表明,本文模型在文本分类任务上也表现出了较好的适用性和有效性。
